Research On Intelligent Shifting Control For Dual Clutch Transmissions Based On Data Driven

Dual Clutch Transmissions(DCT)are widely welcomed by the automotive industry due to their efficient transmission efficiency,low manufacturing cost and good shift characteristics,which are the most promising transmission in China currently.DCT’s shift control strategy has a decisive influence on its shift quality,but its strong nonlinear characteristics lead to complex control.The design of the shift controller still depends on a lot of calibration work.The data-driven control technology can use massive real vehicle data to obtain the optimal shift control law hidden in the data,thereby improving the DCT shift quality.This article takes the 7-speed wet DCT produced by an automobile company as the research object,and carries out the data-driven DCT shift intelligent control research.The main research contents are as follows:(1)Construction of clutch target torque knowledge base based on ensemble learning.The ensemble learning algorithm is used to mine the control law contained in the DCT actual vehicle shift data.Firstly,the shift data under different working conditions is obtained through the actual vehicle shift data collection test,and use the wavelet threshold denoising to preprocess the real vehicle sensor data.Then,design an ensemble learning system for DCT shift control,using preprocessed real vehicle shift data,using DCT system state parameters as input and clutch target torque as output,the nonlinear autoregressive neural network model with external input(NARX)is trained as a submodel of ensemble learning to mine the mapping relationship between the DCT clutch target torque and the ideal shift state.Finally,the RIPPER rule learning algorithm is used to formulate the integration rules of the sub-model,and the knowledge of shift control acquired by ensemble learning is stored in the form of constant neural network weights and sub-model integration rules,the knowledge base of clutch target torque during DCT shift process is constructed,which lays the foundation for knowledge-based intelligent control of DCT shift.(2)DCT system dynamics modeling and simulation.The engine numerical modeling is established based on the engine test data;the torque transmission characteristics of the clutch in different states are analyzed,and the clutch torque transmission model in the DCT system is established;considering the various driving resistances encountered by the vehicle during driving,a driving load model of the vehicle is established.Dynamic analysis and state space modeling of the torque phase and the inertia phase of the DCT shift process are performed respectively.Combined with the above modules,a complete DCT shift dynamics model is constructed.Based on the Matlab / Simulink software platform,the DCT shift process is simulated,and the accuracy of the built model is verified by comparing with the actual vehicle test data under the same working conditions,providing a means for verifying the DCT shifting control strategy.(3)Knowledge-based model predictive control of the DCT shift process.A DCT shift process control method based on a knowledge-driven method and model predictive control(MPC)algorithm is proposed to realize the application of the optimal shift control knowledge.Firstly,according to the actual vehicle state data,the optimal shift control amount under the current operating conditions is extracted from the clutch target torque knowledge base.Then,with the optimization goal of minimizing the clutch rotation speed following error,the MPC torque controller is designed to realize the clutch compensation control,the state prediction equation based on the discrete state space model is derived,and the prediction model of MPC’s shift control system is established.Finally,the objective function of MPC shift control is designed,combined with the range constraints of state variables and control variables,the optimal solution of MPC control variables is transformed into a quadratic programming solution problem,and quadprog function is used to perform rolling optimization for the shift control amount.Based on the Matlab /Simulink simulation platform,the proposed shift control strategy is simulated and verified under upshift and downshift conditions respectively,and compared with the traditional fuzzy shift control strategy and the shift control method based on actual vehicle calibration.(4)Model-free adaptive DCT shift control based on data-driven.In order to improve the adaptability and robustness of the DCT shift controller,a model-free adaptive shift control strategy driven by input / output data of the DCT system is proposed.Combined with the clutch target torque obtained from the actual vehicle data,the shift control knowledge obtained through integrated learning is applied online to the DCT shift control system in the form of data driving.Taking the reduction of longitudinal impact and sliding power during the shift as the optimization goal,the model-free adaptive control(MFAC)algorithm is used to compensate the clutch speed deviation,the MFAC pseudo partial derivative estimation algorithm is designed to update the system’s dynamic linear data model in real time,the MFAC control rate of the DCT shift process is derived,and the shift control amount is calculated in real time based on the clutch target speed and system’s input / output data.Considering the DCT system behavior changes and system internal / external disturbances respectively,the established DCT shift control strategy is simulated at different throttle openings,and compared with the MPC shift simulation results under the same working conditions,and the effectiveness of the DCT model-free adaptive shift control strategy based on data-driven is verified.